Sign In

[Moss Edit] Anime2Real (Anime to Realistic)

3

Updated: Oct 22, 2026

styleanimerealisticanime2real

Not recently loaded

Download

1 variant available

bf16 SafeTensor

Moss_SFT_Edit_A2R_v1_000003000.safetensors

BF16, good balance β€’ 276.81 MB

Verified:

Type
LoRA
Stats

543

3

1

Reviews
Published

Oct 19, 2026

Base Model

MOSS Image

Training
Steps: 3,000
Usage Tips
Strength: 1
Hash
AutoV20AF9A7C6D7
Trigger Words
A2R

Krea2 is still unbeatable so far.

Convert anime image to realistic using moss as edit model, thanks to ostris.

Suggest using with https://civitai.red/models/3012032/moss-edit-anime2real-workflow-anime-to-realistic

Workflow embedded in example images.

ai-toolkit training config file in the Optional Files.

Dependencies:

custom node: https://github.com/ostris/ComfyUI-MOSS-Image-Ostris-Edit

Since ComfyUI has refactored the MossImage code, please use my modified version instead: https://github.com/SurpassHR/ComfyUI-MOSS-Image-Ostris-Edit

Alternatively, you can make modifications yourself:

Edit ComfyUI/custom_nodes/ComfyUI-MOSS-Image-Ostris-Edit/nodes.py

Locate lines 161–162:

Replace the function

def _pe(dit, ids):
    return dit.pe_embedder(ids.unsqueeze(0)).transpose(1, 2).contiguous()

with

def _pe(dit, ids):
    if hasattr(dit, "pe_embedder"):
        return dit.pe_embedder(ids.unsqueeze(0)).transpose(1, 2).contiguous()
    ids_b = ids.unsqueeze(0)
    pe = torch.cat([comfy.ldm.flux.math.rope(ids_b[..., i], d, 10000) for i, d in enumerate(dit.rope_axes_dim)], dim=-3)
    return pe.unsqueeze(1).transpose(1, 2).contiguous()